A technique of motion estimation in video compression is described, the technique including: determining, in one or more reference frames of a video picture, the best full-pixel motion vector F for a block in a current frame of the video picture, wherein m and n are signed numbers and integer multiples of the distance between two adjacent full-pixels; selecting the best half-pixel motion vector candidates from a set of half-pixel motion vectors based on the best full-pixel motion vector; determining the best half-pixel motion vector H; selecting the best quarter-pixel motion vector candidates from a set of quarter-pixel motion vectors based on the best full-pixel motion vector and the best half-pixel motion vector; determining the best quarter-pixel motion vector Q; and determining the best motion vector for the block as BMV.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method of efficiently reducing complexity of sub pixel motion estimation in video compression for limiting degradation, comprising: determining, in one or more reference frames of a video picture, a best full-pixel motion vector F=(m, n) for a block in a current frame of the video picture, wherein m and n are signed numbers and integer multiples of the distance between two adjacent full-pixels; selecting the best half-pixel motion vector candidates from a set of half-pixel motion vectors based on the best full-pixel motion vector; determining the best half-pixel motion vector H; selecting the best quarter-pixel motion vector candidates from a set of quarter-pixel motion vectors based on the best full-pixel motion vector and the best half-pixel motion vector; determining the best quarter-pixel motion vector Q; determining the best motion vector for the block as BMV=F+H+Q; and determining full-pixel motion estimation based on the best motion vector.
Video compression technology for reducing motion estimation complexity. The problem addressed is the computational cost and potential degradation associated with sub-pixel motion estimation. This method efficiently determines a motion vector for a block in a video frame. It begins by finding the optimal full-pixel motion vector, represented as F=(m, n), where m and n are signed integers indicating displacement. Next, a set of potential half-pixel motion vectors is generated and evaluated, with the best half-pixel motion vector, H, being selected based on the previously determined full-pixel vector. Following this, a set of quarter-pixel motion vector candidates is generated and assessed, using both the best full-pixel and best half-pixel vectors to identify the optimal quarter-pixel motion vector, Q. The final best motion vector (BMV) for the block is then calculated by summing the full-pixel, half-pixel, and quarter-pixel vectors (BMV=F+H+Q). Finally, full-pixel motion estimation is performed using this determined best motion vector.
2. The method of claim 1 , wherein the set of half-pixel motion vectors is S1={H1=(2u, 2u), H2=(0, 2u), H3=(−2u, 2u), H4=(2u, 0), H5=(−2u, 0), H6=(2u, −2u), H7=(0, −2u), H8=(−2u, −2u)}; wherein the set of quarter-pixel motion vectors is S2={Q1=(u, u), Q2=(0, u), Q3=(−u, u), Q4=(u, 0), Q5=(−u, 0), Q6=(u, −u), Q7=(0, −u), Q8=(−u, −u)}; and wherein u is a real number greater than zero, and 4u is the distance between two adjacent full-pixels.
3. The method of claim 2 , wherein selecting the best half-pixel motion vector candidates comprising: if the best full-pixel motion vector is a zero vector, selecting all of S1 as the best half-pixel motion vector candidates.
This invention relates to video encoding, specifically improving motion vector prediction by refining half-pixel motion vector candidates. The problem addressed is inefficient motion vector selection, which can degrade compression efficiency and increase computational overhead. The solution involves a method to optimize the selection of half-pixel motion vector candidates based on the characteristics of the best full-pixel motion vector. The method first identifies the best full-pixel motion vector for a current block in a video frame. If this best full-pixel motion vector is a zero vector (indicating no motion), the method selects all half-pixel motion vector candidates from a predefined set S1 as the best candidates. This ensures that when no full-pixel motion is detected, the half-pixel candidates are fully considered, improving accuracy in motion compensation. The predefined set S1 includes half-pixel candidates derived from neighboring blocks or other predictive sources, ensuring comprehensive coverage of potential motion vectors. By conditionally selecting all half-pixel candidates when the full-pixel motion vector is zero, the method avoids unnecessary filtering or pruning of candidates, leading to better motion estimation and improved compression efficiency. This approach reduces computational complexity while maintaining high-quality motion prediction. The method is particularly useful in video encoding standards like H.264, H.265, or AV1, where accurate motion vector prediction is critical for efficient compression.
4. The method of claim 2 , wherein selecting the best half-pixel motion vector candidates comprising: if the best full-pixel motion vector is a non-zero vector and m*n=0, selecting the half-pixel motion vectors in a same and opposite directions of the best full-pixel motion vector as the best half-pixel motion vector candidates.
This invention relates to video encoding, specifically improving motion vector prediction for half-pixel precision. The problem addressed is inefficient half-pixel motion vector selection, which can degrade compression efficiency and visual quality in video encoding. The method involves selecting optimal half-pixel motion vector candidates based on full-pixel motion vector analysis. When the best full-pixel motion vector is non-zero and a specific condition (m*n=0) is met, the process selects half-pixel motion vectors in both the same and opposite directions of the best full-pixel vector. This ensures that the most relevant half-pixel candidates are considered, improving prediction accuracy. The approach leverages spatial correlation between full-pixel and half-pixel motion vectors, reducing redundant computations while maintaining or enhancing encoding performance. By focusing on directionally relevant half-pixel candidates, the method avoids unnecessary evaluations of irrelevant vectors, optimizing both computational efficiency and compression effectiveness. This technique is particularly useful in advanced video codecs where precise motion compensation is critical for achieving high compression ratios with minimal quality loss.
5. The method of claim 2 , wherein selecting the best half-pixel motion vector candidates comprising: if the best full-pixel motion vector is a non-zero vector and m*n≠0, selecting the half-pixel motion vectors in a same trend direction of the best full-pixel motion vector and the half-pixel motion vector in an opposite trend direction of the best full-pixel motion vector as the best half-pixel motion vector candidates; wherein the half-pixel motion vectors in the same trend direction of the best full-pixel motion vector are ( m m 2 u , n n 2 u ) , ( m m 2 u , 0 ) , and ( 0 , n n 2 u ) and the half-pixel motion vector in the opposite trend direction of the best full-pixel motion vector is ( - m m 2 u , - n n 2 u ) , and wherein each of the best half-pixel motion vector candidates belongs to S1.
This invention relates to video encoding, specifically improving motion vector prediction by refining half-pixel motion vector candidates based on the best full-pixel motion vector. The problem addressed is selecting optimal half-pixel motion vectors to enhance prediction accuracy while reducing computational complexity in video compression. The method involves analyzing the best full-pixel motion vector, defined by integer coordinates (m, n). If this vector is non-zero and both m and n are non-zero, the process selects half-pixel motion vectors in two directions: those aligned with the full-pixel vector's trend and one in the opposite direction. The aligned half-pixel candidates are derived by scaling the full-pixel vector components by half their absolute values, producing vectors like (m|m|/2, n|n|/2), (m|m|/2, 0), and (0, n|n|/2). The opposite-direction candidate is (-m|m|/2, -n|n|/2). These candidates are part of a predefined set S1, which likely contains other potential motion vectors for further refinement. This approach ensures that half-pixel candidates are logically connected to the full-pixel vector, improving prediction efficiency by focusing on relevant sub-pixel positions. The method is particularly useful in hybrid video coding standards where motion compensation plays a critical role in reducing redundancy.
6. The method of claim 1 , wherein determining the best half-pixel motion vector comprising: determining the one of the best half-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best half-pixel motion vector.
This invention relates to video encoding and decoding, specifically improving motion estimation by refining motion vectors at half-pixel precision. The problem addressed is accurately determining the best half-pixel motion vector to enhance video compression efficiency and quality. Traditional motion estimation often relies on integer-pixel precision, which can lead to suboptimal predictions and increased residual data. The invention improves upon this by evaluating multiple half-pixel motion vector candidates to select the one that provides the closest match to the current frame's block. The method involves generating multiple half-pixel motion vector candidates, typically derived from integer-pixel motion vectors. Each candidate is evaluated by comparing its corresponding block (from a reference frame) to the block in the current frame. The candidate that yields the best match—typically measured by minimizing a distortion metric such as sum of absolute differences (SAD) or sum of squared differences (SSD)—is selected as the best half-pixel motion vector. This refined motion vector improves prediction accuracy, reducing residual data and enhancing compression efficiency. The technique is particularly useful in video codecs like H.264/AVC and HEVC, where precise motion compensation is critical for high-quality video at lower bitrates.
7. The method of claim 2 , wherein selecting the best quarter-pixel motion vector candidates comprising: determining which of the block corresponding to the best full-pixel motion vector and the block corresponding to the best half-pixel motion vector better matches the block in the current frame.
This invention relates to video encoding, specifically improving motion estimation by refining motion vectors at sub-pixel precision. The problem addressed is the computational inefficiency and accuracy limitations in selecting the best quarter-pixel motion vector candidates during video compression. The method involves comparing two candidate blocks to determine the best quarter-pixel motion vector. The first candidate block is derived from the best full-pixel motion vector, while the second is derived from the best half-pixel motion vector. The method evaluates which of these two blocks provides a closer match to the block in the current frame. This comparison helps refine the motion vector to quarter-pixel precision, improving encoding efficiency without excessive computational overhead. The process begins by identifying the best full-pixel and half-pixel motion vectors through conventional motion estimation techniques. These vectors are then used to generate corresponding blocks in the reference frame. The method then assesses the similarity between these blocks and the current frame's block, selecting the one with the better match. This refined motion vector is used in the final encoding process, enhancing compression quality while maintaining computational feasibility. The approach optimizes the trade-off between accuracy and processing complexity in video encoding.
8. The method of claim 7 , wherein selecting the best quarter-pixel motion vector candidates comprising: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a zero vector, selecting the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the best quarter-pixel motion vector candidate.
In video compression, motion estimation is used to reduce redundancy by predicting blocks in a current frame based on reference frames. A challenge is accurately selecting motion vectors at sub-pixel precision, such as quarter-pixel accuracy, to improve compression efficiency while minimizing computational overhead. This invention addresses this by refining the selection of quarter-pixel motion vector candidates during motion estimation. The method involves evaluating motion vectors at different precision levels—full-pixel, half-pixel, and quarter-pixel—to determine the best match for a block in the current frame. If the block corresponding to the best full-pixel motion vector matches the current block better than other candidates and the best full-pixel motion vector is a zero vector (indicating no motion), the method selects the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the optimal quarter-pixel candidate. This approach leverages the relationship between motion vectors at different precisions to improve accuracy while reducing computational complexity. The technique is particularly useful in video encoding standards where sub-pixel motion estimation is critical for achieving high compression efficiency.
9. The method of claim 7 , wherein selecting the best quarter-pixel motion vector candidates comprising: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a non-zero vector, selecting all of S2 as the best quarter-pixel motion vector candidates.
This invention relates to video encoding, specifically improving motion estimation by refining motion vector selection at the quarter-pixel level. The problem addressed is inefficient motion vector refinement, which can lead to suboptimal compression and visual quality. The method enhances the selection of quarter-pixel motion vector candidates by evaluating the match quality between a block in the current frame and the block corresponding to the best full-pixel motion vector. If the best full-pixel motion vector is non-zero and provides a better match, all candidates from a predefined set S2 are selected as the best quarter-pixel motion vector candidates. This approach ensures that the refinement process leverages the most accurate full-pixel motion vector, improving the efficiency of the quarter-pixel search. The method is part of a broader motion estimation technique that includes initial full-pixel motion vector determination and subsequent refinement at finer granularity. By focusing on the most promising candidates, the technique reduces computational overhead while maintaining or improving encoding performance. The invention is particularly useful in video compression standards where precise motion estimation is critical for achieving high compression ratios and visual fidelity.
10. The method of claim 7 , wherein selecting the best quarter-pixel motion vector candidates comprising: if the block corresponding to the best half-pixel motion vector better matches the block in the current frame, selecting the quarter-pixel motion vectors in the same trend direction of the best half-pixel motion vector, each of which is at an angle less than 90 degrees with the best half-pixel motion vector, as the best quarter-pixel motion vector candidates, wherein each of the best quarter-pixel motion vector candidates belongs to S2.
In video compression, motion estimation is used to reduce redundancy by predicting blocks in a current frame based on motion vectors from a reference frame. A challenge is efficiently selecting optimal quarter-pixel motion vectors to improve prediction accuracy without excessive computation. This invention addresses this by refining the selection of quarter-pixel motion vector candidates based on the best half-pixel motion vector. The method first evaluates whether the block corresponding to the best half-pixel motion vector provides a better match to the current frame block. If so, it selects quarter-pixel motion vectors that align with the trend direction of the best half-pixel motion vector, ensuring each candidate vector forms an angle of less than 90 degrees with the half-pixel vector. These candidates are then classified into a subset S2, which represents the most promising quarter-pixel motion vectors for further refinement. This approach reduces computational overhead by narrowing down candidates to those most likely to yield accurate predictions, improving encoding efficiency while maintaining high-quality video reconstruction.
11. The method of claim 1 , wherein, determining the best quarter-pixel motion vector comprising: determining the one of the best quarter-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best quarter-pixel motion vector.
This invention relates to video encoding and decoding, specifically improving motion estimation accuracy by refining motion vectors at the quarter-pixel level. The problem addressed is the need for more precise motion compensation to enhance video compression efficiency and quality. Traditional motion estimation often relies on integer-pixel or half-pixel precision, which may not capture fine motion details, leading to residual errors and reduced compression performance. The method involves selecting the optimal quarter-pixel motion vector from multiple candidates to minimize prediction error. First, a set of quarter-pixel motion vector candidates is generated, typically by interpolating integer-pixel or half-pixel motion vectors. Each candidate is evaluated by comparing its corresponding block in a reference frame to the block in the current frame. The candidate that yields the closest match, as determined by a similarity metric such as sum of absolute differences (SAD) or sum of squared differences (SSD), is selected as the best quarter-pixel motion vector. This refined motion vector is then used for motion compensation, improving prediction accuracy and reducing residual data. The technique enhances compression efficiency by allowing more precise motion tracking, which is particularly useful for high-resolution or high-motion video content. It can be integrated into existing video codecs like H.264, H.265, or AV1 to improve their performance. The method ensures that the selected motion vector optimally represents the motion between frames, leading to better compression and visual quality.
12. A system of efficiently reducing complexity of sub pixel motion estimation for limiting degradation, comprising: a first determination module adapted for determining, in one or more reference frames of the video picture, a best full-pixel motion vector F=(m, n) for a block in a current frame of a video picture, wherein m and n are signed numbers and integer multiples of the distance between two adjacent full-pixels; a first selection module adapted for selecting the best half-pixel motion vector candidates from a set of half-pixel motion vectors based on the best full-pixel motion vector; a second determination module adapted for determining the best half-pixel motion vector H; a second selection module adapted for selecting the best quarter-pixel motion vector candidates from a set of quarter-pixel motion vectors based on the best full-pixel motion vector and the best half-pixel motion vector; a third determination module adapted for determining the best quarter-pixel motion vector Q; a fourth determination module adapted for determining the best motion vector for the block as BMV=F+H+Q; and a fifth usage module adapted for determining full-pixel motion estimation based on the best motion vector.
The system reduces computational complexity in sub-pixel motion estimation for video processing while minimizing quality degradation. Motion estimation is used to predict motion between video frames, improving compression efficiency. Traditional methods require exhaustive searches at sub-pixel levels (half-pixel and quarter-pixel), which are computationally expensive. This system optimizes the process by hierarchically refining motion vectors at different precision levels. The system first determines the best full-pixel motion vector (F) for a block in the current frame by comparing it to reference frames. Full-pixel vectors are integer multiples of the distance between adjacent pixels. From this, a subset of half-pixel motion vector candidates is selected, and the best half-pixel vector (H) is determined. Similarly, the best quarter-pixel vector (Q) is selected from a subset of candidates based on both the full-pixel and half-pixel vectors. The final motion vector (BMV) is the sum of F, H, and Q. This hierarchical approach reduces the number of candidate vectors evaluated at each sub-pixel level, significantly lowering computational cost while maintaining accuracy. The system then uses the best motion vector for full-pixel motion estimation, ensuring efficient video encoding with minimal quality loss.
13. The motion estimation system of claim 12 , wherein the set of half-pixel motion vectors is S1={H1=(2u, 2u), H2=(0, 2u), H3=(−2u, 2u), H4=(2u, 0), H5=(−2u, 0), H6=(2u, −2u), H7=(0, −2u), H8=(−2u, −2u)}; wherein the set of quarter-pixel motion vectors is S2={Q1=(u, u), Q2=(0, u), Q3=(−u, u), Q4=(u, 0), Q5=(−u, 0), Q6=(u, −u), Q7=(0, −u), Q8=(−u, −u)}; and wherein u is a real number greater than zero, and 4u is the distance between two adjacent full-pixels.
This invention relates to motion estimation in video compression, specifically improving accuracy by refining motion vectors at sub-pixel precision. The system addresses the challenge of efficiently estimating motion between video frames at fractional pixel resolutions, which is critical for reducing bitrate while maintaining visual quality. The system uses predefined sets of half-pixel and quarter-pixel motion vectors to refine motion compensation. The half-pixel motion vectors are defined as S1, consisting of eight vectors with components (2u, 2u), (0, 2u), (-2u, 2u), (2u, 0), (-2u, 0), (2u, -2u), (0, -2u), and (-2u, -2u), where u is a positive real number and 4u represents the distance between adjacent full-pixels. The quarter-pixel motion vectors are defined as S2, consisting of eight vectors with components (u, u), (0, u), (-u, u), (u, 0), (-u, 0), (u, -u), (0, -u), and (-u, -u). These vectors are used to interpolate motion at sub-pixel granularity, enhancing the precision of motion compensation in video encoding. The system likely integrates these vectors into a motion estimation process that first identifies full-pixel motion vectors and then refines them using the half-pixel and quarter-pixel vectors to achieve higher accuracy in motion prediction. This approach improves compression efficiency by reducing residual errors between predicted and actual motion.
14. The motion estimation system of claim 13 , wherein the first selection module selects the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a zero vector, selecting all of S1 as the best half-pixel motion vector candidates.
The motion estimation system is designed to improve video compression by efficiently selecting the best motion vector candidates for half-pixel precision. Motion estimation is a key process in video encoding, where the system predicts motion between frames to reduce redundancy. A common challenge is determining the most accurate motion vectors while minimizing computational complexity. The system addresses this by refining motion vector candidates from full-pixel to half-pixel precision, ensuring high-quality predictions without excessive processing. The system includes a selection module that evaluates full-pixel motion vectors and generates half-pixel candidates based on the best full-pixel vector. If the best full-pixel motion vector is a zero vector—indicating no motion—the module selects all half-pixel candidates from a predefined set (S1) as the best candidates. This approach ensures that when no motion is detected at full-pixel resolution, the system still explores all possible half-pixel refinements to capture subtle motion. The method optimizes accuracy by prioritizing relevant candidates while reducing unnecessary computations. This refinement step is crucial for maintaining high compression efficiency and visual quality in encoded video streams.
15. The motion estimation system of claim 13 , wherein the first selection module selects the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a non-zero vector and m*n=0, selecting the half-pixel motion vectors in a same and opposite directions of the best full-pixel motion vector as the best half-pixel motion vector candidates.
This invention relates to motion estimation in video processing, specifically improving the selection of half-pixel motion vectors for more accurate motion compensation. The problem addressed is the computational inefficiency and potential inaccuracies in traditional motion estimation techniques when refining motion vectors from full-pixel to half-pixel precision. The system includes a motion estimation module that identifies full-pixel motion vectors and a selection module that refines these vectors to half-pixel precision. The selection module evaluates multiple half-pixel candidates around the best full-pixel motion vector. A key feature is the conditional selection of half-pixel candidates based on the characteristics of the best full-pixel motion vector. If the best full-pixel motion vector is non-zero and certain conditions (m*n=0) are met, the system selects half-pixel candidates in both the same and opposite directions of the full-pixel vector. This approach reduces unnecessary computations by focusing on the most relevant half-pixel candidates, improving efficiency and accuracy in motion compensation. The system is particularly useful in video encoding and decoding applications where precise motion estimation is critical for compression efficiency and visual quality. By intelligently narrowing down half-pixel candidates, the invention optimizes the motion estimation process while maintaining high accuracy.
16. The motion estimation system of claim 13 , wherein the first selection module selects the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a non-zero vector and m*n≠0, selecting the half-pixel motion vectors in a same trend direction of the best full-pixel motion vector and the half-pixel motion vector in an opposite trend direction of the best full-pixel motion vector as the best half-pixel motion vector candidates; wherein the half-pixel motion vectors in the same trend direction of the best full-pixel motion vector are m m 2 u , n n 2 v ( m m 2 u , n n 2 u ) , ( m m 2 u , 0 ) m m 2 u · 0 , and 0 · n n 2 v ( 0 , n n 2 u ) and the half-pixel motion vector in the opposite trend direction of the best full-pixel motion vector is ( - m m 2 u , - n n 2 u ) , and wherein each of the best half-pixel motion vector candidates belongs to S1.
The invention relates to motion estimation in video compression, specifically improving the selection of half-pixel motion vector candidates based on the best full-pixel motion vector. The problem addressed is inefficient motion vector refinement, which can lead to suboptimal compression performance. The system selects half-pixel candidates by analyzing the best full-pixel motion vector (m, n). If the full-pixel vector is non-zero and m or n is non-zero, the system selects half-pixel vectors in the same trend direction as the full-pixel vector and one in the opposite direction. The same-direction candidates include vectors scaled by half-pixel precision (e.g., (m|m|*2u, n|n|*2v)) and vectors with one component zero (e.g., (m|m|*2u, 0)). The opposite-direction candidate is (-m|m|*2u, -n|n|*2u). All selected candidates belong to a predefined set S1. This approach ensures that the half-pixel search is focused on the most relevant candidates, improving efficiency and accuracy in motion compensation.
17. The motion estimation system of claim 12 , wherein the second determination module determines the best half-pixel motion vector by: determining the one of the best half-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best half-pixel motion vector.
This invention relates to motion estimation systems used in video compression, specifically improving the accuracy of motion vector determination at the half-pixel level. The system addresses the challenge of efficiently selecting the optimal half-pixel motion vector from multiple candidates to enhance video encoding quality while minimizing computational overhead. The motion estimation system includes a module that evaluates multiple half-pixel motion vector candidates to identify the best match for a block in the current frame. The system first generates candidate half-pixel motion vectors based on integer-pixel motion vectors. A comparison module then assesses each candidate by measuring how closely the corresponding block from a reference frame matches the current frame block. The best half-pixel motion vector is selected as the candidate whose block most closely matches the current frame block, using a matching criterion such as sum of absolute differences (SAD) or sum of squared differences (SSD). This refinement step improves motion compensation accuracy, leading to better compression efficiency and reduced artifacts in encoded video. The system is particularly useful in video codecs like H.264/AVC or HEVC, where precise motion estimation is critical for maintaining high-quality video at lower bitrates. The invention optimizes the trade-off between computational complexity and encoding performance by focusing on half-pixel precision, which is a key step in modern video compression standards.
18. The motion estimation system of claim 12 , wherein the second determination module determines the best half-pixel motion vector by: determining the one of the best half-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best half-pixel motion vector.
The motion estimation system is designed for video encoding, specifically to improve the accuracy of motion vectors used in predictive coding. The system addresses the challenge of efficiently determining the best half-pixel motion vector, which is crucial for reducing prediction errors and improving compression efficiency in video encoding. The system includes a second determination module that evaluates multiple half-pixel motion vector candidates to identify the best one. This module compares each candidate's corresponding block against the block in the current frame, selecting the candidate that provides the closest match. The selected candidate is then designated as the best half-pixel motion vector. This process enhances the precision of motion compensation, leading to better compression and reduced bitrate while maintaining video quality. The system integrates with a broader motion estimation framework that may include integer-pixel motion vector refinement and other optimization techniques to further improve encoding performance. The focus on half-pixel accuracy ensures finer granularity in motion prediction, which is particularly beneficial for scenes with complex motion patterns.
19. The motion estimation system of claim 18 , wherein the second selection module selects the best quarter-pixel motion vector candidates by: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a zero vector, selecting the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the best quarter-pixel motion vector candidate.
This invention relates to motion estimation in video processing, specifically improving the accuracy of motion vector selection for video compression. The system addresses the challenge of efficiently determining the best quarter-pixel motion vector candidates to enhance compression efficiency while reducing computational complexity. The motion estimation system includes a selection module that evaluates motion vector candidates at different precision levels—full-pixel, half-pixel, and quarter-pixel—to identify the most accurate motion vectors for predicting block motion between frames. When the block corresponding to the best full-pixel motion vector matches the current frame block more closely and the best full-pixel motion vector is a zero vector (indicating no motion), the system selects the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the optimal quarter-pixel candidate. This approach refines motion estimation by leveraging higher-precision motion vectors while minimizing unnecessary computations, particularly in scenarios where full-pixel motion vectors are insufficiently accurate. The system ensures improved video compression quality by dynamically adjusting motion vector selection based on frame content and motion characteristics.
20. The motion estimation system of claim 18 , wherein the second selection module selects the best quarter-pixel motion vector candidates by: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a non-zero vector, selecting all of S2 as the best quarter-pixel motion vector candidates.
This invention relates to motion estimation in video processing, specifically improving the accuracy of motion vector selection for video compression. The system addresses the challenge of efficiently determining the best motion vectors at sub-pixel precision, such as quarter-pixel accuracy, while reducing computational complexity. The system includes a motion estimation module that generates full-pixel motion vector candidates and a selection module that refines these candidates to quarter-pixel precision. The selection process involves evaluating the quality of the best full-pixel motion vector. If the block corresponding to the best full-pixel motion vector matches the current frame block more closely and the vector is non-zero, the system selects all quarter-pixel candidates derived from this full-pixel vector as the best candidates. This approach optimizes the search space for sub-pixel refinement, improving encoding efficiency by focusing computational resources on the most promising candidates. The system is particularly useful in video encoding standards like H.264/AVC and HEVC, where accurate motion estimation is critical for compression performance.
21. The motion estimation system of claim 18 , wherein the second selection module selects the best quarter-pixel motion vector candidates comprising: if the block corresponding to the best half-pixel motion vector better matches the block in the current frame, selecting the quarter-pixel motion vectors in the same trend direction of the best half-pixel motion vector, each of which is at an angle less than 90 degrees with the best half-pixel motion vector, as the best quarter-pixel motion vector candidates, wherein each of the best quarter-pixel motion vector candidates belongs to S2.
The invention relates to motion estimation in video compression, specifically improving the selection of quarter-pixel motion vector candidates for more accurate motion compensation. Motion estimation is a key process in video encoding, where the movement of objects between frames is predicted to reduce redundancy. Traditional methods often struggle with fine-grained motion accuracy, particularly at sub-pixel levels like quarter-pixels, leading to inefficient compression and lower video quality. This system enhances motion estimation by refining the selection of quarter-pixel motion vector candidates based on the best half-pixel motion vector. If the block corresponding to the best half-pixel motion vector provides a better match to the current frame block, the system selects quarter-pixel motion vectors that follow the same trend direction as the best half-pixel vector. These candidates are constrained to angles less than 90 degrees relative to the half-pixel vector, ensuring they align closely with the predicted motion path. This approach reduces computational overhead by narrowing the search space while improving accuracy, as it prioritizes candidates that are most likely to yield the best match. The selected candidates belong to a predefined set (S2), which further optimizes the search process. This method improves compression efficiency and video quality by refining motion prediction at sub-pixel levels.
22. The motion estimation system of claim 12 , wherein, the third determination module determines the best quarter-pixel motion vector by: determining the one of the best quarter-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best quarter-pixel motion vector.
This invention relates to motion estimation systems used in video compression, specifically improving the accuracy of motion vector determination at the quarter-pixel level. The system addresses the challenge of efficiently identifying the most accurate motion vector for video encoding, which is critical for reducing bitrate while maintaining visual quality. The invention focuses on refining the selection of quarter-pixel motion vectors, which are used to predict motion between frames at sub-pixel precision. The system includes a module that evaluates multiple quarter-pixel motion vector candidates to determine which one provides the closest match to a corresponding block in the current frame. This involves comparing each candidate's predicted block against the actual block in the current frame to assess similarity. The candidate that yields the best match is selected as the optimal quarter-pixel motion vector. This process enhances motion compensation accuracy, leading to more efficient compression and improved video quality. The invention builds upon a broader motion estimation system that includes modules for generating motion vector candidates, evaluating their performance, and selecting the best candidate based on predefined criteria. The quarter-pixel refinement step ensures that even small motion differences are accurately captured, which is particularly important for high-quality video encoding. The system is designed to operate efficiently within the constraints of video encoding standards, ensuring compatibility with existing compression frameworks.
23. One or more non-transitory computer-readable media for efficiently reducing complexity of sub pixel motion estimation having computer-executable instructions embodied thereon that, when executed by at least one processor, cause at least one processor to: determine, in one or more reference frames of a video picture, a best full-pixel motion vector F=(m, n) for a block in a current frame of the video picture, wherein m and n are signed numbers and integer multiples of the distance between two adjacent full-pixels; select the best half-pixel motion vector candidates from a set of half-pixel motion vectors based on the best full-pixel motion vector; determine the best half-pixel motion vector H; select the best quarter-pixel motion vector candidates from a set of quarter-pixel motion vectors based on the best full-pixel motion vector and the best half-pixel motion vector; determine the best quarter-pixel motion vector Q; determining the best motion vector for the block as BMV=F+H+Q; and determining full-pixel motion estimation based on the best motion vector.
This invention relates to video processing, specifically improving the efficiency of sub-pixel motion estimation in video encoding. The problem addressed is the computational complexity of traditional sub-pixel motion estimation, which involves searching multiple fractional-pixel positions to refine motion vectors. The solution reduces complexity by hierarchically refining motion vectors at half-pixel and quarter-pixel precision levels based on a full-pixel motion vector. The method begins by determining the best full-pixel motion vector (F) for a block in a current video frame, where F is an integer multiple of the distance between adjacent full-pixels. Using this full-pixel vector, a subset of half-pixel motion vector candidates is selected and evaluated to determine the best half-pixel motion vector (H). Similarly, the best quarter-pixel motion vector (Q) is determined from a subset of quarter-pixel candidates, which are selected based on both the full-pixel and half-pixel vectors. The final motion vector (BMV) is computed as the sum of F, H, and Q. This hierarchical refinement reduces the number of candidate vectors evaluated at each sub-pixel level, improving efficiency while maintaining accuracy. The full-pixel motion estimation is then performed using the best motion vector (BMV). This approach minimizes redundant computations by progressively narrowing the search space at each sub-pixel level.
24. The one or more computer readable media of claim 23 , wherein the set of half-pixel motion vectors is S1={H1=(2u, 2u), H2=(0, 2u), H3=(−2u, 2u), H4=(2u, 0), H5=(−2u, 0), H6=(2u, −2u), H7=(0, −2u), H8=(−2u, −2u)}; wherein the set of quarter-pixel motion vectors is S2={Q1=(u, u), Q2=(0, u), Q3=(−u, u), Q4=(u, 0), Q5=(−u, 0), Q6=(u, −u), Q7=(0, −u), Q8=(−u, −u)}; and wherein u is a real number greater than zero, and 4u is the distance between two adjacent full-pixels.
This invention relates to video encoding and decoding, specifically to motion compensation techniques that improve video compression efficiency. The problem addressed is the need for precise motion estimation at sub-pixel resolutions to enhance compression while maintaining visual quality. Traditional motion compensation often relies on full-pixel or half-pixel accuracy, but finer granularity can reduce residual errors and improve compression ratios. The invention defines specific sets of motion vectors for half-pixel and quarter-pixel precision. The half-pixel motion vectors are arranged in a symmetric pattern around a central point, with vectors at positions (2u, 2u), (0, 2u), (-2u, 2u), (2u, 0), (-2u, 0), (2u, -2u), (0, -2u), and (-2u, -2u), where u is a positive real number and 4u represents the distance between adjacent full-pixels. Similarly, the quarter-pixel motion vectors are defined at positions (u, u), (0, u), (-u, u), (u, 0), (-u, 0), (u, -u), (0, -u), and (-u, -u). These predefined sets allow for efficient motion compensation by enabling the encoder to select the most appropriate sub-pixel positions for predicting motion, reducing residual errors and improving compression efficiency. The invention ensures compatibility with existing video coding standards while enhancing performance through finer motion estimation.
25. The one or more computer readable media of claim 24 , wherein execution of the instructions causes the processor to select the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a zero vector, selecting all of S1 as the best half-pixel motion vector candidates.
Video encoding systems use motion vectors to predict and compress video frames by identifying similar blocks between frames. A challenge in this process is efficiently selecting the best motion vector candidates at sub-pixel precision, such as half-pixel accuracy, to improve compression efficiency without excessive computational overhead. This invention addresses this challenge by providing a method for selecting the best half-pixel motion vector candidates in a video encoding system. The method involves evaluating full-pixel motion vectors first, including a zero vector (indicating no motion). If the best full-pixel motion vector is determined to be the zero vector, the method selects all half-pixel candidates from a predefined set S1 as the best candidates for further refinement. This approach ensures that when no significant motion is detected at the full-pixel level, the system still explores all possible half-pixel candidates to capture subtle motion, improving encoding accuracy while maintaining computational efficiency. The method is implemented as part of a video encoding algorithm executed by a processor, with instructions stored on computer-readable media. The selection process is optimized to reduce redundant calculations while ensuring high-quality motion estimation.
26. The one or more computer readable media of claim 24 , wherein execution of the instructions causes the processor to select the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a non-zero vector and m*n=0, selecting the half-pixel motion vectors in a same and opposite directions of the best full-pixel motion vector as the best half-pixel motion vector candidates.
This invention relates to video encoding and decoding, specifically improving motion vector prediction for half-pixel precision. The problem addressed is efficiently selecting optimal half-pixel motion vector candidates during video compression, reducing computational complexity while maintaining encoding quality. The method involves refining motion vector prediction by first determining the best full-pixel motion vector. If this vector is non-zero and certain conditions are met (where m and n are parameters related to motion vector components), the system selects half-pixel motion vectors in both the same and opposite directions of the full-pixel vector as the best candidates. This approach leverages spatial correlation in video frames to predict motion more accurately with fewer computations. The technique is part of a broader video encoding system that processes video frames by dividing them into blocks, predicting motion between frames, and encoding residual data. The half-pixel refinement step improves prediction accuracy by considering sub-pixel motion, which is crucial for smooth video playback. The method ensures efficient candidate selection by focusing only on relevant half-pixel positions, reducing the search space while maintaining high-quality motion compensation. This is particularly useful in real-time video applications where processing efficiency is critical.
27. The one or more computer readable media of claim 24 , wherein execution of the instructions causes the processor to select the best half-pixel motion vector candidates by: if the best full-pixel motion vector is a non-zero vector and m*n≠0, selecting the half-pixel motion vectors in a same trend direction of the best full-pixel motion vector and the half-pixel motion vector in an opposite trend direction of the best full-pixel motion vector as the best half-pixel motion vector candidates; wherein the half-pixel motion vectors in the same trend direction of the best full-pixel motion vector are m m 2 u , n n 2 v ( m m 2 u , n n 2 u ) , ( m m 2 u , 0 ) m m 2 u · 0 , and 0 · n n 2 v ( 0 , n n 2 u ) and the half-pixel motion vector in the opposite trend direction of the best full-pixel motion vector is ( - m m 2 u , - n n 2 u ) , and wherein each of the best half-pixel motion vector candidates belongs to S1.
In video compression, motion estimation is used to reduce redundancy by predicting frames based on motion vectors. A challenge is efficiently selecting the best half-pixel motion vector candidates to improve prediction accuracy while minimizing computational complexity. This invention addresses this by refining the selection process for half-pixel motion vectors based on the best full-pixel motion vector. When the best full-pixel motion vector is non-zero and its components (m, n) are non-zero, the method selects half-pixel motion vectors in the same trend direction as the full-pixel vector and one in the opposite trend direction. The same-trend candidates include vectors derived from the full-pixel vector's components, scaled by half-pixel precision, such as (m|m|/2 u, n|n|/2 v), (m|m|/2 u, 0), and (0, n|n|/2 v). The opposite-trend candidate is (-m|m|/2 u, -n|n|/2 v). These candidates are part of a predefined set (S1) to ensure efficient evaluation. The approach optimizes motion vector refinement by focusing on relevant half-pixel candidates, improving prediction accuracy while reducing computational overhead.
28. The one or more computer readable media of claim 23 , wherein execution of the instructions causes the processor to determine the best half-pixel motion vector by: determining the one of the best half-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best half-pixel motion vector.
The invention relates to video encoding and decoding, specifically improving motion estimation by refining motion vectors at the half-pixel level. In video compression, motion estimation predicts pixel movement between frames to reduce redundancy. However, traditional methods often use integer-pixel precision, which can lead to inaccuracies in motion representation. This invention addresses the problem by refining motion vectors to half-pixel precision, improving prediction accuracy and compression efficiency. The system determines the best half-pixel motion vector by evaluating multiple half-pixel motion vector candidates. Each candidate represents a potential motion vector refined to half-pixel precision. The system compares each candidate's corresponding block with the block in the current frame. The candidate whose block most closely matches the current frame block is selected as the best half-pixel motion vector. This refinement process enhances motion prediction accuracy, leading to better compression and reduced artifacts in encoded video. The invention builds on prior techniques by incorporating half-pixel refinement into the motion estimation process. By selecting the best candidate based on block matching, it ensures optimal motion representation at a higher precision level. This approach is particularly useful in high-definition video encoding, where finer motion details are critical for maintaining quality. The method improves compression efficiency while preserving visual fidelity.
29. The one or more computer readable media of claim 24 , wherein execution of the instructions causes the processor to select the best quarter-pixel motion vector candidates by: determining which of the block corresponding to the best full-pixel motion vector and the block corresponding to the best half-pixel motion vector better matches the block in the current frame.
In the field of video compression, motion estimation is used to reduce redundancy by predicting motion between frames. A challenge in this process is efficiently selecting the best motion vectors at sub-pixel precision, such as quarter-pixel accuracy, to improve compression efficiency without excessive computational overhead. This invention addresses this challenge by providing a method for selecting the best quarter-pixel motion vector candidates. The method involves comparing two blocks: one corresponding to the best full-pixel motion vector and another corresponding to the best half-pixel motion vector. The block that better matches the block in the current frame is selected as the reference for further quarter-pixel refinement. This approach ensures that the initial motion vector selection is optimized before proceeding to finer precision levels, improving accuracy while reducing unnecessary computations. The technique is particularly useful in video encoding standards that support sub-pixel motion compensation, such as H.264/AVC and HEVC, where efficient motion estimation is critical for achieving high compression ratios. By leveraging the best full-pixel and half-pixel candidates, the method enhances the overall efficiency of the motion estimation process.
30. The one or more computer readable media of claim 29 , wherein execution of the instructions causes the processor to select the best quarter-pixel motion vector candidates by: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a zero vector, selecting the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the best quarter-pixel motion vector candidate.
In video encoding, motion estimation is used to reduce redundancy by predicting blocks in a current frame based on reference frames. A challenge is efficiently selecting the best motion vector candidates at sub-pixel precision, such as quarter-pixel accuracy, to improve compression efficiency without excessive computational overhead. This invention relates to a method for selecting the best quarter-pixel motion vector candidates during motion estimation. The process involves evaluating full-pixel and half-pixel motion vectors to refine the search for the optimal quarter-pixel motion vector. Specifically, if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a zero vector (indicating no motion), the system selects the quarter-pixel motion vector in the opposite direction of the best half-pixel motion vector as the best quarter-pixel motion vector candidate. This approach leverages existing motion vector information to guide the selection of quarter-pixel candidates, reducing computational complexity while maintaining encoding efficiency. The method is implemented via executable instructions stored on computer-readable media, executed by a processor to perform the motion vector selection.
31. The one or more computer readable media of claim 29 , wherein execution of the instructions causes the processor to select the best quarter-pixel motion vector candidates by: if the block corresponding to the best full-pixel motion vector better matches the block in the current frame and the best full-pixel motion vector is a non-zero vector, selecting all of S2 as the best quarter-pixel motion vector candidates.
This invention relates to video encoding and decoding, specifically improving motion vector prediction by refining quarter-pixel motion vector candidates. The problem addressed is the computational inefficiency and accuracy limitations in selecting optimal motion vectors for video compression, particularly when refining from full-pixel to quarter-pixel precision. The method involves analyzing motion vectors to improve video encoding efficiency. A full-pixel motion vector is first determined for a block in a video frame. If the block corresponding to this full-pixel motion vector closely matches the block in the current frame and the full-pixel motion vector is non-zero, all quarter-pixel candidates from a predefined set (S2) are selected as the best candidates for further refinement. This approach reduces the number of candidate vectors evaluated, improving processing speed while maintaining encoding quality. The predefined set S2 includes quarter-pixel motion vectors derived from the full-pixel motion vector, ensuring that only the most relevant candidates are considered. By focusing on the most promising candidates, the method optimizes the encoding process without sacrificing accuracy. This technique is particularly useful in video compression standards where motion estimation is a critical step in achieving efficient encoding.
32. The one or more computer readable media of claim 29 , wherein execution of the instructions causes the processor to select the best quarter-pixel motion vector candidates by: if the block corresponding to the best half-pixel motion vector better matches the block in the current frame, selecting the quarter-pixel motion vectors in the same trend direction of the best half-pixel motion vector, each of which is at an angle less than 90 degrees with the best half-pixel motion vector, as the best quarter-pixel motion vector candidates, wherein each of the best quarter-pixel motion vector candidates belongs to S2.
This invention relates to video encoding and decoding, specifically improving motion vector prediction for quarter-pixel precision. The problem addressed is efficiently selecting optimal quarter-pixel motion vector candidates to reduce computational complexity while maintaining encoding quality. In video compression, motion vectors estimate the movement of blocks between frames, and higher precision (like quarter-pixel) improves accuracy but increases processing overhead. The invention optimizes this by first evaluating half-pixel motion vectors. If the block corresponding to the best half-pixel vector matches the current frame block better than the integer-pixel vector, the system selects quarter-pixel candidates in the same trend direction as the best half-pixel vector. These candidates must form angles less than 90 degrees with the half-pixel vector and belong to a predefined subset (S2). This approach narrows down the search space for quarter-pixel candidates, reducing computational effort while maintaining prediction accuracy. The method leverages directional trends from half-pixel vectors to guide quarter-pixel candidate selection, ensuring efficient and effective motion estimation.
33. The one or more computer readable media of claim 23 , wherein, execution of the instructions causes the processor to determine the best quarter-pixel motion vector by: determining the one of the best quarter-pixel motion vector candidates, whose corresponding block best matches the block in the current frame, as the best quarter-pixel motion vector.
The computer figures out the most accurate motion of objects between video frames by comparing several slightly different motion options and picking the one that produces the closest matching image.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
September 26, 2014
April 18, 2017
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.